After almost two years of hype following the release of ChatGPT, it is worth asking whether the large language model (LLM) sector is overvalued.
It comes at a time when AI vendors are struggling to innovate. Even OpenAI, which recently achieved a $150 billion valuation, hasn’t confirmed the release of GPT-5. While GPT-4o was an interesting addition to its line up in terms of multimodal, OpenAI o1 is arguably the vendor’s most underwhelming release yet
Though OpenAI claims the model is its best at Chain of Thought reasoning, many people expected what was known as ‘Project Strawberry’ to be a precursor to GPT-5, rather than a rough research model that even Sam Altman admits “Seems more impressive on first use than it does after you spend more time with it.”
Across the board, AI vendors seem to be struggling, not only with regard to justifying the need for generative AI as a solution but also for differentiating themselves from other providers, particularly when so many solutions like Anthropic, Google, Microsoft, X, and Meta exist in the market.
Key Takeaways
- The large language model (LLM) sector may be overvalued, with many AI vendors struggling to innovate.
- Even OpenAI has faced criticism for not delivering GPT-5 as anticipated, despite a $150 billion valuation.
- Generative AI tools like ChatGPT often provide inaccurate or biased outputs, requiring rigorous fact-checking.
- The AI industry is also facing legal challenges, particularly around copyright infringement claims.
- Some experts argue that while generative AI is overhyped, it has significant potential in specific, everyday use cases.
Why Generative AI is Overvalued
One of the biggest issues surrounding generative AI is the level of money that’s being spent on the market, with little tangible return.
Back in June, Goldman Sachs released a newsletter noting that despite tech giants and other entities planning to invest $1 trillion in AI (PDF), along with data centers, chips, the power grid and other infrastructure, “this spending has little to show for it so far.”
As part of the company’s report, Daron Acemoglu, Institute Professor at MIT estimated that only a quarter of AI-exposed tasks will be cost-effective to automate within the next 10 years.
The truth about modern LLMs is that in many use cases, they can often more trouble than they’re worth. Even top chatbots like ChatGPT with GPT-4o will hallucinate and produce verifiably false outputs. This means they need to be rigorously fact checked if used as search assistants or content creation tools.
Ameesh Divatia, CEO and co-founder of data security vendor Baffle, told Techopedia:
“It is unclear whether GenAI can provide meaningful customer value as the issues with hallucinations/biased responses, “stealing” copyrighted work, and a lack of reasoning haven’t been resolved.
GenAI has been referred to as a stochastic parrot for a reason since it lies often and with authority!”
Other key researchers in the industry, such as Yann LeCunn, Meta’s chief AI scientist, have spoken openly about the limitations of generative AI, noting that these models have a “very limited understanding of logic… Do not understand the physical world, do not have persistent memory, cannot reason in any reasonable definition of the term and cannot plan…hierarchically.”
Generative AI’s Sword of Damocles: The Copyright Debate
The problems with LLMs go well beyond their technical limitations. There are numerous allegations and lawsuits against leading AI vendors like OpenAI, Stability, Midjourney and Runway for copyright infringement.
In a multi-billion dollar lawsuit, The New York Times alleges that OpenAI used millions of the publication’s copyrighted news articles to train its proprietary AI models without permission or payment.
A similar case alleges that Stability, Midjourney, and Runway had used visual artists’ work to train their LLM without permission.
While it’s impossible to know whether such cases will play out in court, it does raise questions over what would happen to the industry if a court found AI vendors liable for violating the copyright and obligated them to pay damages.
It also raises questions over whether downstream companies that use an offending model could also face legal liabilities for use of copyrighted content.
For example, if a court finds Midjourney liable for use of protected artists’ works, would that mean that a company that used an AI-generated image as part of its marketing would also face legal and financial liabilities? At this stage it’s hard to know because there isn’t much in the way of legal precedent.
In this sense, copyright debate looms over the generative AI market like the sword of Damocles, with the potential to disrupt the progress of AI vendors and adoptees alike.
Counterview: Is AI Undervalued?
Of course there are those who will argue that AI is undervalued. The core argument here is that AI is in its infancy, and while there has been lots of investment, the AI revolution is just getting started.
In fact, companies like PWC estimate that AI could contribute up to $15.7 trillion to the economy in 2030.
To put that in perspective, that’s more than the output of China and India combined.
So while hype about AGI being around the corner might be a bit too far, there are smaller use cases where generative AI can be a force multiplier.
Jeremiah Stone, CTO at SnapLogic, told Techopedia:
“There’s a lot of overhype in consumer generative AI with sensational stories of GenAI replacing workers, but there’s a different application among enterprises where companies are using generative AI to enhance and assist their employees [which] is very undervalued.”
Perhaps it is in these everyday use cases of making workflows that are a little bit faster where generative AI has the biggest value to provide, rather than in reinventing the wheel or replacing human workers completely.
Philip Gjørup, co-founder at Nord Comms, told Techopedia:
“I really don’t believe generative AI is overvalued. If anything, it’s packed with potential to spark creativity and drive innovation.
Take a look at companies like Grok, Corcel.io, Venice.ai, Dipply.ai, and Bidmindlabs.ai. They’re great examples of how AI can amplify what humans are capable of across a variety of fields. That’s exactly why venture capitalists are heavily investing in these kinds of startups, and why Nvidia, for instance, has really increased its AI spending.”
The Bottom Line
Although trillion dollar investments may seem a little much for a technology that’s still very rough around the edges, there is enough potential for AI to contribute significantly to the global economy over the next few years.
This won’t come from replicating or replacing human intelligence, but in helping to streamline everyday tasks humans perform everyday in the workplace and at home.